- Email: [email protected]
Cost-Effectiveness of Rosuvastatin Compared with Other Statins from a Managed Care Perspective
Blackwell Science, LtdOxford, UKVHEValue in Health1098-30152005 ISPOR86618628Original ArticleFormulary Decisions in the Statin ClassBenner et al.
Volume 8 • Number 6 • 2005 VALUE IN HEALTH
Cost-Effectiveness of Rosuvastatin Compared with Other Statins from a Managed Care Perspective Joshua S. Benner, PharmD, ScD,1 Timothy W. Smith, BA,1 David Klingman, PhD,1 Jonothan C. Tierce, CPhil,1 C. Daniel Mullins, PhD,2 Ned Pethick, MBA,3 John C. O’Donnell, PhD3 1
ValueMedics Research, LLC, Arlington, VA, USA; 2University of Maryland, Baltimore, MD, USA; 3AstraZeneca LP, Wilmington, DE, USA
ABST R ACT
Objective: The objective of this study was to identify the most cost-effective statin or combination of statins, from the perspective of a managed care payer. Methods: A decision-analytic model compared the cost-effectiveness of titration to goal with atorvastatin, fluvastatin, lovastatin, pravastatin, rosuvastatin, and simvastatin in patients with elevated low-density lipoprotein cholesterol (LDL-C). Effectiveness measures included the percentage change from baseline LDL-C and high-density lipoprotein cholesterol (HDL-C), and the percentage of patients achieving National Cholesterol Education Program (NCEP) Second Adult Treatment Panel (ATP II) LDL-C goals. Direct medical costs were calculated based on drug, physician, and laboratory resource use, multiplied by wholesale acquisition costs for drugs and the 2005 Medicare reimbursement rates for services. A Monte Carlo simulation tested the sensitivity of results to model efficacy inputs. Results: In the base-case analysis, rosuvastatin dominated atorvastatin, pravastatin, and simvastatin. Generic
lovastatin dominated fluvastatin. The incremental (absolute) reduction in LDL-C, increase in HDL-C, and increase in patients to goal with rosuvastatin compared with lovastatin were 16%, 3%, and 27%, respectively. Incremental costs per additional 1% reduction in LDL-C, 1% increase in HDL-C, and patient to goal with rosuvastatin versus lovastatin were $8, $41, and $436, respectively. A wide variety of assumptions were assessed and Monte Carlo sensitivity analyses were conducted. Findings were most sensitive to the cost of lovastatin. Conclusion: Rosuvastatin dominates atorvastatin, pravastatin, and simvastatin because it is more effective and less costly, and it may be considered cost-effective compared with generic lovastatin. The most cost-effective two-statin formulary contained lovastatin and rosuvastatin. Keywords: cost-effectiveness, hydroxymethylglutarylCoA reductase inhibitors, hypercholesterolemia, managed care.
Background
lipid-lowering in the US population. Rates of cholesterol testing and use of lipid-lowering medications remain low, particularly among high-risk patients [11–13]. Among those who are treated, low-density lipoprotein cholesterol (LDL-C) goals are achieved in fewer than half of patients [14], even after statin therapy is titrated [15]. The 1999–2000 National Health and Nutrition Examination Survey (NHANES) revealed that awareness, treatment, and control of hypercholesterolemia among US adults have changed little in the past decade [16]. Clinical and economic considerations in the management of dyslipidemia have changed substantially since the most recent NHANES data were collected. The availability of generic lovastatin has substantially lowered the cost to initiate therapy within this class. Ezetimibe, a cholesterol absorption inhibitor, is a new adjunct to statin therapy in patients who fail to achieve goals on monotherapy. Finally, rosu-
Cardiovascular disease (CVD) accounts for more than 930,000 deaths annually in the United States, including more than 500,000 deaths from coronary heart disease (CHD) [1]. The annual cost of CVD is $368 billion, including $227 billion in direct medical expenditures [1]. Since 1994, numerous large clinical trials have shown that 3-hydroxy-3-methylglutaryl-coenzyme A reductase inhibitors (statins) significantly reduce the incidence of cardiovascular morbidity and mortality [2–8]. Although the rate of statin use has risen in recent years [9,10], there is substantial unmet need for Address correspondence to: Joshua S. Benner, Principal, ValueMedics Research, LLC, 300 N, Washington St., Suite 303, Falls Church, VA 22046, USA. E-mail: [email protected] 10.1111/j.1524-4733.2005.00055.x © ISPOR
1098-3015/05/618
618–628
618
619
Formulary Decisions in the Statin Class vastatin became the sixth statin available in the United States after its approval by the Food and Drug Administration in July 2003. Long-term outcomes data are not yet available for rosuvastatin, but in a 6-week comparative clinical trial, rosuvastatin 10 mg to 40 mg lowered LDL-C to levels <100 mg/dL in a significantly higher proportion of patients compared with groups receiving equivalent doses of atorvastatin or equivalent and higher doses of simvastatin and pravastatin [17]. The general aim of this study was to assist managed care decision-makers as they reevaluate statin formularies in light of these new therapeutic alternatives. The specific objective was to identify the most cost-effective statin(s) from a payer perspective, based on the currently available clinical data and guidelines.
Methods A cost-effectiveness analysis (CEA) was designed to inform statin selection in a managed care setting. A decision-analytic model was constructed that compared the treatment costs and effectiveness of each statin. Consistent with the Academy of Managed Care Pharmacy (AMCP) Format for Formulary Submissions Version 2.0 [18], the analysis employed the payer perspective, hence only direct medical costs, and time horizons of one and 3 years (a lifetime analysis was not conducted because longterm clinical data were not yet available for all of the treatments used in the model). The statins and their starting and maximum dosages included atorvastatin 10 mg to 80 mg, fluvastatin 40 mg to 80 mg, generic lovastatin 20 mg to 40 mg, pravastatin 20 mg to 40 mg, rosuvastatin 10 mg to 40 mg, and simvastatin 20 mg to 80 mg. Doses were titrated to target LDL-C levels specified in the Second Adult Treatment Panel (ATP II) of the National Cholesterol Education Program (NCEP) or to the maximum dosage specified above. Patient-level effectiveness parameters, derived from clinical trial
data and the published literature, included percentage reduction in LDL-C, percentage increase in high-density lipoprotein cholesterol (HDL-C), and achievement of ATP II goal. Resource-use parameters were based on treatment guidelines issued by the Third NCEP Adult Treatment Panel (ATP III) and included statin costs, physician visits, and laboratory tests. Unit costs were based on Medicare reimbursement rates and the wholesale acquisition cost (WAC) of statins. All key parameters were subjected to extensive sensitivity analyses, including Monte Carlo simulation. Details regarding each of these elements of the analysis are given below. The conceptual framework of the model is depicted in Figure 1. The model was constructed using Excel 2002 (Microsoft Corp.) and @Risk 4.5 (Palisade Corp.). The base-case analysis compared the cost-effectiveness of titration to LDL-C goal with each statin in patients with baseline LDL-C ≥ 160 mg/dL and <250 mg/dL. To facilitate comparisons between statins, the incremental cost-effectiveness ratio (ICER) for each drug was the difference in cost between the drug and the next less costly alternative divided by the difference in effectiveness between the drug and the next less costly alternative [19]. Interventions which are more effective and less costly than the alternative (dominant) [19] or which have a lower ICER than the alternative (extended dominance) are preferred [20]. The undominated products comprise the “efficient frontier” [19], which is also referred to in this article as the optimal statin formulary.
Patient Population The analysis employed patient-level data from two multicenter, Phase III, randomized clinical trials. One compared rosuvastatin with atorvastatin [21] and the other compared rosuvastatin with pravastatin and simvastatin [22], both over 52 weeks. These trials had virtually identical inclusion criteria and protocols, and are the only two long-term trials
Figure 1 Model flow—base case. All patients initiated statin therapy on the recommended starting dose and were titrated at their physicians’ discretion until they achieved their ATP II goal or until they reached the highest available dose. Patients who achieved goal continued on maintenance therapy at the current dose. Patients who failed to achieve goal continued on maintenance therapy at the highest available dose.
Benner et al.
620 that compare rosuvastatin to other statins. In the Olsson study [21], 412 patients from 45 centers in northern Europe with LDL-C 160 mg/dL to 249 mg/dL were randomized to rosuvastatin 5 mg, rosuvastatin 10 mg, or atorvastatin 10 mg. After 12 weeks of treatment, dosages could be sequentially doubled up to 80 mg if the patient was not at his/her ATP II LDL-C goal. In the Brown study [22], 477 patients from 43 centers in the United States who met identical inclusion criteria were randomized to rosuvastatin 5 mg, rosuvastatin 10 mg, pravastatin 20 mg, or simvastatin 20 mg. After 12 weeks of fixed-dose therapy, dosages were sequentially doubled to a maximum of 80 mg for rosuvastatin and simvastatin and 40 mg for pravastatin according to investigator discretion if the patient did not meet his/her ATP II LDL-C goal. The 52-week observed data from the Olsson [21] and Brown [22] trials were pooled. Because rosuvastatin was substantially more effective in the European (Olsson [21]) study than in the US (Brown [22]) study, a multivariate linear regression model was fit to the pooled database of rosuvastatin patients from both trials. The main effect for the trial variable (adjusting for clinical and demographic characteristics that were associated with effectiveness) was used to weight the 52-week lipid levels among all Olsson [21] subjects downward to make them similar to those of Brown [22] subjects. The range of effectiveness used in sensitivity analysis was broad enough to cover potential error in the adjustment weights. Subjects who did not complete the 52-week follow-up period were excluded for three reasons. First, the objective of this analysis was to estimate effectiveness and costs in patients treated for at least 1 year. Second, discontinuation rates were low (<20%) and similar across treatment arms, so dropping these patients could be done without biasing the results for any particular drug [21,22]. Third, exclusion of these subjects avoided potentially inaccurate imputation of costs and effectiveness in the base-case analysis. In sensitivity analyses, missing data were imputed for the full intent-to-treat population. Because 10 mg is the recommended starting dose for rosuvastatin [2], and 75% of all rosuvastatin sold is in the form of 10 mg tablets [24], patients randomized to 5 mg were also excluded from this analysis. Based on the characteristics of included trial participants (atorvastatin n = 116; pravastatin n = 95; rosuvastatin n = 202; simvastatin n = 102), the hypothetical cohort in the model was 40% male and had a mean age of 58 years and a mean baseline LDL of 189 mg/dL. Approximately
23% of the patients were considered “high risk” by ATP II criteria; another 35% were at moderately high risk.
Statin Effectiveness Effectiveness measures included the percentage change in LDL-C, the percentage change in HDL-C, and the percentage of patients achieving ATP II LDL-C goal (Table 1). ATP II goals were used for this analysis because comparative 1-year effectiveness data for rosuvastatin is at present only available from trials in which subjects were titrated to ATP II targets. For the base-case analysis, effectiveness of atorvastatin, pravastatin, rosuvastatin and simvastatin was based on measurements at 52 weeks (after titration to goal) in the pooled trial cohort described above. The modest incremental effectiveness in the “base-case” versus “no titration” scenarios may have been because titrations were done after randomization at the physician’s discretion, so patients who were titrated may have been farther from goal or treatment-resistant. Five patients in the trials were titrated to rosuvastatin 80 mg; however, after completion of the clinical program, a decision was made by the manufacturer not to pursue marketing approval of the 80 mg dose. Accordingly, these patients’ 52-week efficacy data were adjusted by carrying forward their last observations at the 40 mg dose. Sensitivity analyses were conducted in which lower and upper 95% confidence limits were used instead of means for effectiveness parameters. These ranges were generally consistent with the results of other long-term clinical trials [2–5,7,25] and product labels. Although fluvastatin and lovastatin were not used in the Olsson [21] and Brown [22] trials, these low-cost statins were included in the CEA to make it more relevant to managed care decision-makers. Effectiveness of these drugs was estimated based on the principle of dose equivalence. Previous research suggests that fluvastatin is half as potent as pravastatin, but the two drugs achieve comparable changes in LDL-C and HDL-C over 52 weeks when given in equipotent doses [26,27]. Thus, because pravastatin was given over the 20 mg to 40 mg dose range in the Brown trial [22], it was assumed that the results would be comparable to fluvastatin over the 40 mg to 80 mg dose range. Similarly, lovastatin 20 mg to 40 mg is approximately equivalent to pravastatin 20 mg to 40 mg [26]. The assumption that fluvastatin 40 mg to 80 mg and lovastatin 20 mg to 40 mg were equivalent to pravastatin 20 mg to 40 mg in terms of LDL-C reduction, HDL-C improvement, and ATP II goal achievement
621
Formulary Decisions in the Statin Class Table 1 Statin effectiveness parameters under base-case and alternative scenarios Statin Atorvastatin 10–80 mg Base-case (95% CI) No titration With ezetimibe Reduced compliance Fluvastatin 40–80 mg Base-case (95% CI) No titration With ezetimibe Reduced compliance Lovastatin 20–40 mg Base-case (95% CI) No titration With ezetimibe Reduced compliance Pravastatin 20–40 mg Base-case (95% CI) No titration With ezetimibe Reduced compliance Rosuvastatin 10–40 mg Base-case (95% CI) No titration With ezetimibe Reduced compliance Simvastatin 20–80 mg Base-case (95% CI) No titration With ezetimibe Reduced compliance
% Reduction in LDL-C
% Increase in HDL-C
% Achieving ATP II goal
38 (36, 40) 36 42 30
0.9 (−1.3, 3.0) 8.5 9.5 —
80 (73, 87) 70 94 52
30 (28, 33) 28 36 24
4.4 (2.1, 6.6) 8.3 9.8 —
60 (50, 70) 56 86 46
30 (28, 33) 28 36 24
4.4 (2.1, 6.6) 8.3 9.8 —
60 (50, 70) 56 86 46
30 (28, 33) 28 36 24
4.4 (2.1, 6.6) 8.3 9.8 —
60 (50, 70) 56 86 46
46 (44, 48) 49 50 36
7.3 (5.4, 9.1) 11.3 11.7 —
87 (83, 92) 89 99 67
37 (34, 39) 37 42 29
6.1 (3.5, 8.7) 9.2 10.1 —
73 (64, 81) 72 91 53
Upper and lower bounds of the 95% CI were used in one-way sensitivity analyses. ATP II, National Cholesterol Education Program, Second Adult Treatment Panel; CI, confidence interval; HDL-C, high-density lipoprotein cholesterol; LDL-C, lowdensity lipoprotein cholesterol; —, not available.
is also consistent with data from long-term trials of those drugs [6,8,25].
Costs Consistent with the managed care perspective, direct medical costs of lipid-lowering therapy were included in this analysis. First-year total cost of treatment was estimated based on statin use, physician office visits, and laboratory monitoring (Table 2). In the base-case analysis, it was assumed that physician visits and lab tests were performed according to ATP III guidelines [28]. The model cohort therefore incurred two physician visits, two lipid panels, one liver function test (LFT), and one creatine kinase (CK) test before initiating therapy, and follow-up visits for titration every 8 weeks until the maximum dose or LDL-C goal was reached. Thereafter, follow-up intervals were reduced to every 6 months. Lipid tests were assumed to occur at each physician visit, and repeat LFT and CK assessments were done at week 8 only. The number of physician visits and lipid tests within the first year of therapy was therefore determined by the number of titrations required to achieve LDL-C goal. Average 2005 Medicare reimbursement rates were used as a measure of the costs of physician vis-
its and laboratory procedures [29,30]. A range of five Common Procedural Terminology, Fourth Edition (CPT-4) codes (99211–99215) for physician office visits were used to estimate this cost. The second highest visit level (second most complicated case) was selected for the base case [31], and the minimum and maximum values in the range were used in sensitivity analyses. CPT-4 codes 80061, 80076, and 82550 were chosen to represent the costs of lipid panels, LFT, and CK, respectively. The minimum and maximum observed values in the range were used in the sensitivity analyses for each parameter. The WAC (80% of average wholesale price [AWP]) for the largest-selling package size as of January 19, 2005 was used to determine the daily cost of each statin. The exception was lovastatin, a multisource generic product, where the average WAC across all manufacturers was used [32]. Because there is great variability in net acquisition cost due to unpublished contractual arrangements and rebates, sensitivity analyses probed a range of WAC less 25% to WAC plus 25% (equivalent to AWP). Threshold analyses identified the percent change in acquisition cost for each product that was required to alter the base-case efficient frontier.
Benner et al.
622 Table 2 Cost parameters (in 2004 US dollars) Parameter Physician visit (CPT 99214) Lipid panel (CPT 80061) Liver function test (CPT 80076) Creatine kinase (CPT 82550) Lipid-lowering drugs (per day) Atorvastatin 10 mg 20/40/80 mg Ezetimibe 10 mg Fluvastatin 40 mg 80 mg Lovastatin 20 mg 40 mg Pravastatin 20 mg 40 mg Rosuvastatin 10/20/40 mg Simvastatin 20/40/80 mg
Base-case
Low value
High value
82.62 18.19 11.02 9.08
21.60 13.69 6.41 8.62
120.14 18.75 11.42 9.10
2.17 3.15
1.63 2.36
2.71 3.94
2.28
1.71
2.85
1.75 2.25
1.31 1.69
2.19 2.81
1.34 2.44
1.01 1.83
1.68 3.05
2.79 4.10
2.09 3.08
3.49 5.13
2.36
1.77
2.95
3.97
2.98
4.96
Compliance with statin therapy was assumed to be 100% in the base-case analysis, because this study pertained only to patients who completed 52 weeks of follow-up, and because differences in adherence between the statins have not been documented. In a separate scenario, the sensitivity of base-case results to reduced compliance was tested.
Adverse Events and Long-Term Outcomes Adverse events were not calculated in the model because available evidence suggests that treatmentlimiting event rates do not differ significantly between the statins [21,22,33]. Moreover, the average cost of adverse events would be low, because most events resolve after discontinuation of the drug [2–8]. Although long-term postmarketing safety data are not yet available for rosuvastatin, cases of rhabdomyolysis are very rare and similar in frequency to the other marketed statins [34]. Also excluded were costs for future clinical outcomes such as myocardial infarction, stroke, coronary artery bypass grafting or percutaneous transluminal coronary angioplasty, even though statin therapy has been shown to reduce the frequency of these procedures [2–8]. Excluding these potential cost offsets is consistent with the short-term time frame of the analysis and gives a more conservative estimate of the cost-effectiveness of statin therapy, particularly among the most effective statins [35].
Sensitivity Analyses and Alternate Scenarios To determine the robustness of base-case findings, alternate assumptions about each variable in the
model were tested. In addition to one-way sensitivity analyses on each variable, a Monte Carlo simulation was used to vary all effectiveness inputs at once [36,37]. The best-fitting continuous probability distribution for each drug’s effectiveness was mathematically identified from the patient-level clinical trial data. Lovastatin and fluvastatin were assigned their own distributions (with the same parameters as pravastatin, as explained earlier), to allow their effectiveness to vary independently from pravastatin in the simulation. Although costs varied with the number of titrations required to reach the goal, statin prices were not varied in the Monte Carlo simulation, as they are constant within a health plan. Values from each probability distribution were randomly selected as the model generated 1000 populations of 10,000 patients each. Costeffectiveness acceptability curves were plotted to display the optimal alternatives as a function of willingness-to-pay for incremental effectiveness [37]. In addition, the impacts of five alternate scenarios that may be reflective of actual practice were also studied. The first scenario assumed that patients were not titrated to goal, but instead completed the year on the initial statin dose. For this analysis, effectiveness at the initial dose was carried forward from the 12-week visit in the clinical trials. In a second scenario, it was assumed that ezetimibe 10 mg was added to statin therapy in lieu of titration for patients who did not achieve LDL-C goal. Patients were assumed to incur an additional visit and lipid panel when initiating ezetimibe, as well as
623
Formulary Decisions in the Statin Class Table 3 1-year and 3-year base-case cost per 1% reduction in LDL-C
Strategy
Average cost ($) 1-year 3-year
Lovastatin Fluvastatin Rosuvastatin Atorvastatin Pravastatin Simvastatin
1280 1242 1326 1401 1778 1955
3120 3177 3476 3676 4846 5280
Incremental cost ($) 1-year 3-year — Dominated 118 Dominated Dominated Dominated
— Dominated 356 Dominated Dominated Dominated
Incremental % ↓ LDL-C 1-year 3-year
Average % ↓ LDL-C 30 30 46 38 30 37
— Dominated 16 Dominated Dominated Dominated
— Dominated 16 Dominated Dominated Dominated
Incremental costeffectiveness ratio ($/1% ↓ LDL-C) 1-year 3-year — Dominated 8 Dominated Dominated Dominated
— Dominated 23 Dominated Dominated Dominated
Cost-effectiveness ratios were calculated before cost and effectiveness estimates were rounded. LDL-C, low-density lipoprotein cholesterol; —, not applicable.
the cost of the drug. The additional effectiveness with ezetimibe add-on therapy was estimated based on efficacy data for patients who have ezetimibe added to an ongoing HMG-CoA therapy regimen which is contained on the product label [38]. The third scenario explored the potential impact of noncompliance on cost-effectiveness. Statin, physician, and laboratory utilization were reduced by 16% based on data from a 12-month study of statin compliance among managed care enrollees [39]. Effectiveness in terms of LDL-C reduction was adjusted downward by 20% in this scenario [40]. Although the Olsson [21] and Brown [22] trials were not powered to detect differences in effectiveness across NCEP ATP II risk subgroups, scenario four examined cost-effectiveness by NCEP ATP II risk group. Finally, the fifth scenario examined costeffectiveness over a 3-year time horizon, consistent with the recommendations of the AMCP Format for Formulary submissions [18].
Results
Base-Case Analysis In the base-case analysis, the mean reductions in LDL-C for rosuvastatin, atorvastatin, and simvastatin were 46%, 38%, and 37%, respectively. Pravastatin, lovastatin, and fluvastatin had mean LDL-C reductions of 30%. Generic lovastatin had
the lowest first-year cost of therapy ($1208), followed by fluvastatin ($1242), rosuvastatin ($1326), atorvastatin ($1401), pravastatin ($1778), and simvastatin ($1955). When the drugs were compared for incremental cost-effectiveness, lovastatin dominated fluvastatin, whereas rosuvastatin dominated atorvastatin, simvastatin, and pravastatin (Tables 3, 4, and 5). The first-year incremental cost of rosuvastatin was thus $118 compared with lovastatin, or $8 per additional 1% reduction in LDLC, $436 per additional patient to ATP II goal, and $41 per additional 1% increase in HDL-C. Atorvastatin 10 mg had a lower acquisition cost than rosuvastatin 10 mg, yet the total first-year cost of rosuvastatin was lower because of greater effectiveness at the starting dose, which prevented titrations in all but 16% of patients. Moreover, the flat price of rosuvastatin minimized the economic impact when patients were titrated. Conversely, 37% of patients treated with atorvastatin required at least one titration. In addition to more frequent physician monitoring and laboratory tests, drug costs increased substantially at higher doses of atorvastatin.
Sensitivity Analyses Base-case results were most sensitive to the acquisition costs of rosuvastatin, atorvastatin, and lovastatin. A decrease in the price of rosuvastatin by 14%
Table 4 1-year and 3-year base-case cost per patient to ATP II goal*
Strategy Lovastatin Fluvastatin Rosuvastatin Atorvastatin Pravastatin Simvastatin
Average cost ($) 1-year 3-year 12,080,450 12,424,554 13,264,305 14,005,258 17,780,272 19,547,992
31,197,003 31,767,946 34,756,905 36,763,638 48,457,949 52,801,642
Incremental cost ($) 1-year 3-year — Dominated 1,183,855 Dominated Dominated Dominated
— Dominated 3,559,902 Dominated Dominated Dominated
Average patients to ATP II goal 6,000 6,000 8,713 8,017 6,000 7,255
*Assuming 10,000 patients treated with each statin. Cost-effectiveness ratios were calculated before cost and effectiveness estimates were rounded. ATP II, National Cholesterol Education Program, Second Adult Treatment Panel; —, not applicable.
Incremental patients to ATP II goal 1-year 3-year — Dominated 2,713 Dominated Dominated Dominated
— Dominated 2,713 Dominated Dominated Dominated
Incremental costeffectiveness ratio ($/patient to ATP II goal) 1-year 3-year — Dominated 436 Dominated Dominated Dominated
— Dominated 1,312 Dominated Dominated Dominated
Benner et al.
624 Table 5 1-year and 3-year base-case cost per 1% increase in HDL-C
Strategy Lovastatin Fluvastatin Rosuvastatin Atorvastatin Pravastatin Simvastatin
Average cost ($) 1-year 3-year 1208 1242 1326 1401 1778 1955
3120 3177 3476 3676 4846 5280
Incremental cost ($) 1-year 3-year — Dominated 118 Dominated Dominated Dominated
— Dominated 356 Dominated Dominated Dominated
Incremental % ↑ HDL-C 1-year 3-year
Average % ↑ HDL-C 4.4 4.4 7.3 0.9 4.4 6.1
— Dominated 2.9 Dominated Dominated Dominated
— Dominated 2.9 Dominated Dominated Dominated
Incremental costeffectiveness ratio ($/1% ↑ HDL-C) 1-year 3-year — Dominated 41 Dominated Dominated Dominated
— Dominated 122 Dominated Dominated Dominated
Cost-effectiveness ratios were calculated before cost and effectiveness estimates were rounded. HDL-C, high-density lipoprotein cholesterol; —, not applicable.
or more made it the least costly and most effective alternative, dominating all other statins. Conversely, for prices up to 30% greater than its WAC, rosuvastatin remained the only branded statin on the efficient frontier. Nevertheless, the ICER at WAC plus 30% was $24 per additional 1% decrease in LDL (three times the base case ICER), compared with lovastatin. When the price of atorvastatin was discounted 15% below WAC, it joined the efficient frontier with an ICER of $7 per additional 1% decrease in LDL compared with lovastatin. The ICER for rosuvastatin in that scenario was $8 compared with atorvastatin. When atorvastatin was discounted 24% or more, it became the least costly statin, and the ICER for rosuvastatin was $18 per additional 1% decrease in LDL. When the average price of lovastatin was replaced with the lowest available price (36% less than the base-case price), its total cost in year 1 was $956, and the ICER for rosuvastatin was $24 per additional 1% decrease in LDL compared with lovastatin. Pravastatin and simvastatin remained dominated until their base-case prices were discounted 45% and 49%, respectively. Results were moderately sensitive to assumptions about HDL-C improvement. Moderate increases in the base-case HDL-C benefits of simvastatin and fluvastatin were sufficient for these products to avoid being dominated. For example, if fluvastatin improved HDL-C by 6.6%, it joined the efficient frontier, and the ICER for rosuvastatin increased threefold ($123 per additional 1% increase in HDLC compared with fluvastatin). When simvastatin was assumed to increase HDL-C by 8.7%, it joined the efficient frontier, but its relatively high price yielded an ICER of $443 per additional 1% improvement in HDL-C, compared with rosuvastatin. Results were insensitive to the cost of physician visits and lab procedures, as well as to variations in effectiveness over the 95% confidence limits for LDL-C reduction. At its upper confidence limit for ATP II goal achievement, atorvastatin was the most
effective statin. Nevertheless, its higher price and small incremental benefit relative to rosuvastatin yielded a high ICER ($27,313 per additional patient to goal versus rosuvastatin). Results did not differ substantially from the base case when costs and benefits were imputed for patients who discontinued therapy during the study year. Figure 2 provides the cost-effectiveness acceptability frontier for each outcome of interest, based on 1000 simulated populations of 10,000 patients each. In all cases, the frontier included only lovastatin and rosuvastatin. The intersection of the lines illustrates the willingness-to-pay threshold where the optimal drug of choice changes from lovastatin to rosuvastatin. These willingness-to-pay thresholds are approximately $8, $434, and $40 for a 1% reduction in LDL-C, one additional patient to ATP II goal, and a 1% increase in HDL-C, respectively.
Alternative Scenarios When patients were assumed to remain at their respective starting doses and 12-week effectiveness persisted for the full year, lovastatin remained the least costly alternative ($933), followed by fluvastatin ($1083), atorvastatin ($1236), rosuvastatin ($1305), pravastatin ($1463), and simvastatin ($1894). Although atorvastatin was less costly than rosuvastatin in this scenario, atorvastatin was still dominated in the extended sense because rosuvastatin had a lower incremental cost per additional unit of benefit compared with lovastatin. When patients received add-on therapy with ezetimibe instead of statin titration, lovastatin remained the least costly strategy and rosuvastatin was the most effective, with an ICER of $10 per additional 1% reduction in LDL-C. Rosuvastatin dominated fluvastatin, atorvastatin, pravastatin, and simvastatin. Similarly, when resource use and effectiveness were adjusted for compliance rates representative of actual practice, the results were virtually identical to the base-case.
625
Formulary Decisions in the Statin Class A. Cost-effectiveness acceptability curves for LDL-C reduction 100.00 Rosuvastatin
% of Time Optimal
80.00 60.00 Threshold = $7.60 40.00 20.00 Lovastatin 0.00 $0.00
$5.00 $10.00 Willingness To Pay
$15.00
B. Cost-effectiveness acceptability curves for HDL-C increase 100.00
% of Time Optimal
Rosuvastatin 80.00 60.00 Threshold = $40.32 40.00 20.00
Discussion
Lovastatin 0.00 $0.00
$20.00
$40.00 $60.00 Willingness To Pay
$80.00
C. Cost-effectiveness acceptability curves for ATP II goal achievement 100.00 Rosuvastatin
% of Time Optimal
80.00
60.00 Threshold = $434.06 40.00
20.00 Lovastatin 0.00 $0.00
$200.00
$400.00 $600.00 Willingness To Pay
ditional 1% reduction in LDL for rosuvastatin versus lovastatin was $6, $10, and $7 in the high-risk, moderately high risk, and low-risk groups, respectively. These ratios reflect the fact that LDL reductions were greatest among the high-risk patients (51% for rosuvastatin and 32% for lovastatin), followed by low-risk patients (46% and 29%), and moderately high risk patients (43% and 31%). Finally, when the base-case scenario was evaluated using a 3-year time horizon, total costs increased to reflect longer-term statin use (Tables 3, 4, and 5). Nevertheless, effectiveness was the same as in the 1-year analysis because, under recommended monitoring and titration intervals, all titrations occur within the first year of treatment. Thus, the ICERs in the 3-year analysis may be interpreted as the cost to maintain a given level of effectiveness for 3 years.
$800.00
Figure 2 Results of Monte Carlo simulation. (A) Cost-effectiveness acceptability curves for LDL-C reduction. The threshold is interpreted as the point at which each therapy is optimal 50% of the time based on an expressed willingness to pay for each incremental 1% reduction in LDL-C. (B) Cost-effectiveness acceptability curves for HDL-C increase. The threshold is interpreted as the point at which each therapy is optimal 50% of the time based on an expressed willingness to pay for each incremental 1% increase in HDL-C. (C) Costeffectiveness acceptability curves for ATP II goal achievement. The threshold is interpreted as the point at which each therapy is optimal 50% of the time based on an expressed willingness to pay for each incremental patient to ATP II goal.
In the subanalysis of cost-effectiveness by NCEP ATP II risk group, the base-case results remained relatively stable, but the cost-effectiveness ratio for rosuvastatin was most favorable for patients at high risk of coronary events. The incremental cost per ad-
This CEA compared currently available statins in patients with dyslipidemia from the perspective of a managed care payer. When patients were titrated to ATP II goals, the most effective strategy was rosuvastatin over the 10 mg to 40 mg dose range. Atorvastatin, simvastatin, and pravastatin were less effective and more costly than rosuvastatin. Lovastatin over the 20 mg to 40 mg range was the least costly strategy, followed by fluvastatin over the 40 mg to 80 mg range. Compared with lovastatin, the incremental cost of rosuvastatin was $118 per patient in the initial year of therapy, or $8 per additional 1% reduction in LDL-C, $41 per additional 1% increase in HDL-C, and $436 per additional patient to ATP II goal. These ratios increased by as much as three times as the price of generic lovastatin was reduced from the average of all available manufacturers to the lowest published price. These findings have potentially important implications for managed care decision-makers. Under the base-case and each alternative scenario, atorvastatin, pravastatin, and simvastatin were always dominated. Thus, depending on actual acquisition prices, payers may achieve substantial cost savings and greater effectiveness by using rosuvastatin instead of these agents. In the Monte Carlo simulation of 1000 populations, the cost-effectiveness acceptability curves favored generic lovastatin and rosuvastatin as the optimal formulary. This conclusion was consistent with that of Morrison and Glassberg, who asserted that a cost-effective approach to lipid-lowering therapy would be to treat low-risk patients and those with lower base-
626 line LDL-C levels with the least costly statin, while using the most effective drug in high-risk patients and those with high baseline LDL-C levels [41]. The present study extends those theoretical findings based on currently available pricing and clinical data. Moreover, the results of this analysis can be used to calculate the trade-offs inherent in selecting a different formulary. For example, a payer might consider placing generic lovastatin and simvastatin on formulary until long-term outcomes and safety data become available for rosuvastatin. Assuming 10,000 patients treated with each product, in year 1 the choice of simvastatin would cost about $6.3 million more than rosuvastatin, and 1458 fewer patients would achieve ATP II LDL-C goals. There remains an important question of whether the additional cost of rosuvastatin compared with lovastatin ($436 per additional patient to LDL-C goal) constitutes good value for money [42]. To put this in a decision-making context, the analysis may be recalculated in the absence of rosuvastatin. The next most effective alternative is atorvastatin, which is currently on many formularies. Under base-case assumptions, atorvastatin, in the absence of rosuvastatin, has an incremental cost per patient to goal of $954, compared with lovastatin. Thus, rosuvastatin is able to get more patients to goal than atorvastatin, and at a 54% lower incremental cost per patient to goal. This analysis was unique in its analytic approach, because it was neither purely trial-based nor purely model-based. Typical trial-based analyses are subject to the limitations of protocol-driven resource use, which does not apply to real-world populations. Conversely, purely model-based analyses require that multiple sources of data be synthesized and extrapolated to common endpoints in ways that enhance external validity at the expense of internal validity. Given that no long-term outcomes or safety studies are yet available for rosuvastatin or ezetimibe, we developed a simple model based on clinical guidelines for the management of dyslipidemia patients, but limited the base-case time horizon and effectiveness endpoints to those used in the available clinical trials. This approach thus placed the available clinical data in the context of a real-world formulary decision, thereby balancing the tradeoffs between internal and external validity. It may well be a useful method for evaluating new products in other therapeutic areas. These findings should be interpreted in light of some limitations. Because the objective was to identify the optimal combination of statins for a managed care formulary, several nonstatin medications
Benner et al. for dyslipidemia (e.g., niacin, fibrates, and bile sequestrants) were excluded from analysis. These products generally have a secondary role in therapy, are relatively inexpensive, and were not expected to influence the relative cost-effectiveness of products in the statin class. Their exclusion from this analysis should not be construed to mean that they are not cost-effective agents for modifying serum lipids. The effectiveness of atorvastatin, pravastatin, rosuvastatin, and simvastatin were estimated based on data from two Phase III trials. Fluvastatin and lovastatin were assumed to be equivalent to pravastatin over the dose ranges specified, because long-term trials have not compared fluvastatin and lovastatin with rosuvastatin. Although the resulting estimates for each drug are consistent with numerous other trials and product labels, and although these estimates were varied in sensitivity analyses, the results may not apply to patient populations with different characteristics than those in the Olsson [21] and Brown [22] trials. The subgroup analysis by NCEP ATP II risk group should be interpreted with caution, as the trials were not powered to detect differences in effectiveness across risk strata. Costs were estimated using published WAC for drugs and Medicare reimbursement rates for physician visits and laboratory procedures. These costs may not reflect the actual costs negotiated by managed care organizations, but wide variation in these parameters were probed to find relevant thresholds that changed the optimal formulary. Costs associated with adverse drug events and nonstudy medications were excluded, but these were not expected to differ across treatment groups, and therefore would not have affected the incremental analysis. This analysis had a 1- to 3-year time horizon, and therefore examined surrogate endpoints instead of long-term outcomes such as cost per life-year gained. Although long-term head-to-head outcomes studies comparing the agents in this analysis would be a preferable data source [43], decisions must be made based on currently available data. As such, this model informs payers of the trade-offs inherent in covering one statin (or combination of statins) over another. Moreover, studying surrogate endpoints in this analysis yielded conservative estimates of the incremental effectiveness of rosuvastatin. Estimating long-term events prevented and associated cost offsets based on recent clinical trial data [35] would increase the incremental effectiveness and decrease the incremental cost of rosuvastatin. Finally, achievement of ATP III LDL-C goals could not be measured because patients were titrated to ATP II goals in the Olsson [23] and Brown [24] tri-
627
Formulary Decisions in the Statin Class als. Nevertheless, a recent head-to-head trial of ATP III goal achievement after 6 weeks of treatment with rosuvastatin, atorvastatin, simvastatin, or pravastatin resulted in incrementally higher goal achievement with rosuvastatin [44]. The magnitude of the incremental effectiveness was similar to the inputs employed in the present analysis. In conclusion, the findings of this analysis indicate that rosuvastatin is less costly and more effective than atorvastatin, pravastatin, and simvastatin over the dose ranges for which comparative data are available. Compared with fluvastatin and pravastatin, generic lovastatin is cost-saving and about as effective. Therefore, generic lovastatin and rosuvastatin comprise the optimal two-statin formulary. Compared with generic lovastatin, rosuvastatin offers substantial additional LDL-C reduction, HDL-C improvement, and NCEP goal achievement at a reasonable additional cost, depending on the payer’s willingness to pay for these benefits. Formulary decisions based on these results should be revisited periodically, as new pricing, outcomes and safety data become available.
7
8
9
10
11
12 Source of financial support: This research was sponsored by AstraZeneca.
13
References 1 American Heart Association. 2004 Heart and Stroke Statistical Update. Dallas, TX: American Heart Association, 2003. 2 The Scandinavian Simvastatin Survival Study Group. Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S). Lancet 1994;344:1383–9. 3 Shepherd J, Cobbe SM, Ford I, et al. for the West of Scotland Coronary Prevention Study Group. Prevention of coronary heart disease in men with hypercholesterolemia. N Engl J Med 1995;333: 1301–7. 4 Sacks FM, Pfeffer MA, Moye LA, et al. for the Cholesterol and Recurrent Events Trial Investigators. The effect of pravastatin on coronary events after myocardial infarction in patients with average cholesterol levels. N Engl J Med 1996;335: 1001–9. 5 The Long-Term Intervention with Pravastatin in Ischaemic Disease (LIPID) Study Group. Prevention of cardiovascular events and death with pravastatin in patients with coronary heart disease and a broad range of initial cholesterol levels. N Engl J Med 1998;339:1349–57. 6 Downs JR, Clearfield M, Weis S, et al. for the AFCAPS/TexCAPS Research Group. Primary pre-
14
15
16
17
18
vention of acute coronary events with lovastatin in men and women with average cholesterol levels: results of AFCAPS/TexCAPS. JAMA 1998;279: 1615–22. Heart Protection Study Collaborative Group. MRC/BHF heart protection study of cholesterol lowering with simvistatin in 20,536 high-risk individuals: a randomised placebo-controlled trial. Lancet 2002;360:7–22. Serruys PWJC, de Feyter P, Macaya C, et al. Fluvastatin for prevention of cardiac events following successful first percutaneous coronary intervention. JAMA 2002;287:3215–22. Jackevicius CA, Anderson GM, Leiter L, Tu JV. Use of the statins in patients after acute myocardial infarction: does evidence change practice? Arch Intern Med 2001;161:183–8. Lemaitre RN, Furberg CD, Newman AB, et al. Time trends in the use of cholesterol-lowering agents in older adults. Arch Intern Med 1998;158: 1761–8. Hoerger TJ, Bala MV, Bray JW, et al. Treatment patterns and distribution of low-density lipoprotein cholesterol levels in treatment-eligible United States adults. Am J Cardiol 1998;82:61–5. Stafford RS, Blumenthal D, Pasternak RC. Variations in cholesterol management practices of U.S. physicians. J Am Coll Cardiol 1997;29:139–46. Fonarow GC, French WJ, Parsons LS, et al. Use of lipid-lowering medications at discharge in patients with acute myocardial infarction: data from the National Registry of Myocardial Infarction 3. Circulation 2001;103:38–44. Pearson TA, Laurora I, Chu H, Kafonek S. The lipid treatment assessment project (L-TAP): a multicenter survey to evaluate the percentages of dyslipidemic patients receiving lipid-lowering therapy and achieving low-density lipoprotein cholesterol goals. Arch Intern Med 2000;160:459–67. Foley KA, Simpson RJ, Crouse JR, et al. Effectiveness of statin titration on low-density lipoprotein cholesterol goal attainment in patients at high risk of atherogenic events. Am J Cardiol 2003;92:79– 81. Ford ES, Mokdad AH, Giles WH, Mensah GA. Serum total cholesterol concentrations and awareness, treatment, and control of hypercholesterolemia among US adults: findings from the National Health and Nutrition Examination Survey, 1999– 2000. Circulation 2003;107:2185–9. McKenney JM, Jones PH, Adamczyk MA, et al. Comparison of the efficacy of rosuvastatin vs. atorvastatin, simvastatin, and pravastatin in achieving lipid goals: results from the STELLAR trial. Curr Med Res Opin 2003;19:689–98. Academy of Managed Care Pharmacy. Format for Formulary Submissions, Version 2.0. Alexandria, VA: Academy of Managed Care Pharmacy, 2002.
Benner et al.
628 19 Garber AM, Weinstein MC, Torrance GW, Kamlet MS. Theoretical foundations of cost-effectiveness analysis. In: Gold MR, Siegel JE, Russel LB, Weinstein MC, eds., Cost-Effectiveness in Health and Medicine. New York: Oxford University Press, 1996. 20 Siegel JE, Weinstein MC, Torrance GW. Reporting cost-effectiveness studies and results. In: Gold MR, Siegel JE, Russel LB, Weinstein MC, eds., CostEffectiveness in Health and Medicine. New York: Oxford University Press, 1996. 21 Olsson AG, Istad H, Luurila O, et al. Effects of rosuvastatin and atorvastatin compared over 52 weeks of treatment in patients with hypercholesterolemia. Am Heart J 2002;144:1044–51. 22 Brown WV, Bays HE, Hassman DR, et al. Efficacy and safety of rosuvastatin compared with pravastatin and simvastatin in patients with hypercholesterolemia: a randomized, double-blind, 52-week trial. Am Heart J 2002;144:1036–43. 23 AstraZeneca LP. Crestor® (rosuvastatin calcium) prescribing information. August 2003. 24 First Data Bank. Data on file, AstraZeneca. 25 Davidson M, McKenney J, Stein E, et al. Comparison of one-year efficacy and safety of atorvastatin vs. lovastatin in primary hypercholesterolemia. Am J Cardiol 1997;79:1475–81. 26 Roberts WC. The rule of 5 and the rule of 7 in lipid-lowering by statin drugs [Editorial]. Am J Cardiol 1997;80:106–7. 27 Andrews TC, Ballantyne CM, Hsia JA, Kramer JH. Achieving and maintaining National Cholesterol Education Program low-density lipoprotein cholesterol goals with five statins. Am J Med 2001;111: 185–91. 28 Third Report of the Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III). National Cholesterol Education Program, National Heart Lung and Blood Institute November 19, 2002. WWW document: Available from: http://www.nhlbi.nih.gov/guidelines/cholesterol/ atp3_rpt.htm [Accessed June 4, 2004]. 29 2004 National Physician Fee Schedule Relative Value File. WWW document: Available from: http://www.cms.hhs.gov/providers/pufdownload/ rvudown.asp [Accessed June 4, 2004]. 30 2004 Clinical Diagnostic Laboratory Fee Schedule. WWW document: Available from: http://www.
31
32 33 34 35
36
37
38 39 40
41 42 43 44
cms.hhs.gov/providers/pufdownload/clfdown.asp [Accessed June 4, 2004]. Ganz DA, Kuntz KA, Jacobson GA, Avorn J. Costeffectiveness of 3-hydroxy-3-methylglutaryl Coenzyme A reductase inhibitor therapy in older patients with myocardial infarction. Ann Intern Med 2000;132:780–7. First DataBank National Drug Data File. Data on file, Astra Zeneca. Thompson PD, Clarkson P, Karas RH. Statinassociated myopathy. JAMA 2003;289:1681–90. Olsson GO. Safety and efficacy of rosuvastatin [Letter]. Lancet 2004;364:135. Cannon CP, Braunwald E, McCabe CH, et al. Intensive versus moderate lipid lowering with statins after acute coronary syndromes. N Engl J Med 2004;350:1495–504. Doubilet P, Begg CB, Weinstein MC, et al. Probabilistic sensitivity analysis using Monte Carlo simulation: a practical approach. Med Decis Making 1985;5:157–77. Briggs AH, O’Brien BJ, Blackhouse G. Thinking outside the box: recent advances in the analysis and presentation of uncertainty in costeffectiveness studies. Annu Rev Public Health 2002;23:377–401. Zetia® (ezetimibe) Prescribing Information. March 2003, Merck/Schering-Plough Pharmaceuticals, Inc., Table 2. Fitterman LK, Williams SA. A descriptive review of statin utilization patterns. Am Heart J 2002;144: 905–6. Frolkis JP, Pearce GL, Nambi V, et al. Statins do not meet expectation for lowering low-density lipoprotein cholesterol levels when used in clinical practice. Am J Med 2002;113:625–9. Morrison A, Glassberg H. Determinants of the cost-effectiveness of statins. J Manag Care Pharm 2003;9:544–51. Ohsfeldt RL. Challenges in evaluating the costeffectiveness of statins [Letter]. J Manag Care Pharm 2004;10:81–2. Hay JW. Evaluating the cost-effectiveness of statins [Letter]. J Manag Care Pharm 2004;10:79–81. Jones PH, Davidson MH, Stein EA, et al. Comparison of the efficacy and safety of rosuvastatin versus atorvastatin, simvastatin, and pravastatin across doses (STELLAR Trial). Am J Cardiol 2003;93:152–60.
Volume 8 • Number 6 • 2005 VALUE IN HEALTH
Cost-Effectiveness of Rosuvastatin Compared with Other Statins from a Managed Care Perspective Joshua S. Benner, PharmD, ScD,1 Timothy W. Smith, BA,1 David Klingman, PhD,1 Jonothan C. Tierce, CPhil,1 C. Daniel Mullins, PhD,2 Ned Pethick, MBA,3 John C. O’Donnell, PhD3 1
ValueMedics Research, LLC, Arlington, VA, USA; 2University of Maryland, Baltimore, MD, USA; 3AstraZeneca LP, Wilmington, DE, USA
ABST R ACT
Objective: The objective of this study was to identify the most cost-effective statin or combination of statins, from the perspective of a managed care payer. Methods: A decision-analytic model compared the cost-effectiveness of titration to goal with atorvastatin, fluvastatin, lovastatin, pravastatin, rosuvastatin, and simvastatin in patients with elevated low-density lipoprotein cholesterol (LDL-C). Effectiveness measures included the percentage change from baseline LDL-C and high-density lipoprotein cholesterol (HDL-C), and the percentage of patients achieving National Cholesterol Education Program (NCEP) Second Adult Treatment Panel (ATP II) LDL-C goals. Direct medical costs were calculated based on drug, physician, and laboratory resource use, multiplied by wholesale acquisition costs for drugs and the 2005 Medicare reimbursement rates for services. A Monte Carlo simulation tested the sensitivity of results to model efficacy inputs. Results: In the base-case analysis, rosuvastatin dominated atorvastatin, pravastatin, and simvastatin. Generic
lovastatin dominated fluvastatin. The incremental (absolute) reduction in LDL-C, increase in HDL-C, and increase in patients to goal with rosuvastatin compared with lovastatin were 16%, 3%, and 27%, respectively. Incremental costs per additional 1% reduction in LDL-C, 1% increase in HDL-C, and patient to goal with rosuvastatin versus lovastatin were $8, $41, and $436, respectively. A wide variety of assumptions were assessed and Monte Carlo sensitivity analyses were conducted. Findings were most sensitive to the cost of lovastatin. Conclusion: Rosuvastatin dominates atorvastatin, pravastatin, and simvastatin because it is more effective and less costly, and it may be considered cost-effective compared with generic lovastatin. The most cost-effective two-statin formulary contained lovastatin and rosuvastatin. Keywords: cost-effectiveness, hydroxymethylglutarylCoA reductase inhibitors, hypercholesterolemia, managed care.
Background
lipid-lowering in the US population. Rates of cholesterol testing and use of lipid-lowering medications remain low, particularly among high-risk patients [11–13]. Among those who are treated, low-density lipoprotein cholesterol (LDL-C) goals are achieved in fewer than half of patients [14], even after statin therapy is titrated [15]. The 1999–2000 National Health and Nutrition Examination Survey (NHANES) revealed that awareness, treatment, and control of hypercholesterolemia among US adults have changed little in the past decade [16]. Clinical and economic considerations in the management of dyslipidemia have changed substantially since the most recent NHANES data were collected. The availability of generic lovastatin has substantially lowered the cost to initiate therapy within this class. Ezetimibe, a cholesterol absorption inhibitor, is a new adjunct to statin therapy in patients who fail to achieve goals on monotherapy. Finally, rosu-
Cardiovascular disease (CVD) accounts for more than 930,000 deaths annually in the United States, including more than 500,000 deaths from coronary heart disease (CHD) [1]. The annual cost of CVD is $368 billion, including $227 billion in direct medical expenditures [1]. Since 1994, numerous large clinical trials have shown that 3-hydroxy-3-methylglutaryl-coenzyme A reductase inhibitors (statins) significantly reduce the incidence of cardiovascular morbidity and mortality [2–8]. Although the rate of statin use has risen in recent years [9,10], there is substantial unmet need for Address correspondence to: Joshua S. Benner, Principal, ValueMedics Research, LLC, 300 N, Washington St., Suite 303, Falls Church, VA 22046, USA. E-mail: [email protected] 10.1111/j.1524-4733.2005.00055.x © ISPOR
1098-3015/05/618
618–628
618
619
Formulary Decisions in the Statin Class vastatin became the sixth statin available in the United States after its approval by the Food and Drug Administration in July 2003. Long-term outcomes data are not yet available for rosuvastatin, but in a 6-week comparative clinical trial, rosuvastatin 10 mg to 40 mg lowered LDL-C to levels <100 mg/dL in a significantly higher proportion of patients compared with groups receiving equivalent doses of atorvastatin or equivalent and higher doses of simvastatin and pravastatin [17]. The general aim of this study was to assist managed care decision-makers as they reevaluate statin formularies in light of these new therapeutic alternatives. The specific objective was to identify the most cost-effective statin(s) from a payer perspective, based on the currently available clinical data and guidelines.
Methods A cost-effectiveness analysis (CEA) was designed to inform statin selection in a managed care setting. A decision-analytic model was constructed that compared the treatment costs and effectiveness of each statin. Consistent with the Academy of Managed Care Pharmacy (AMCP) Format for Formulary Submissions Version 2.0 [18], the analysis employed the payer perspective, hence only direct medical costs, and time horizons of one and 3 years (a lifetime analysis was not conducted because longterm clinical data were not yet available for all of the treatments used in the model). The statins and their starting and maximum dosages included atorvastatin 10 mg to 80 mg, fluvastatin 40 mg to 80 mg, generic lovastatin 20 mg to 40 mg, pravastatin 20 mg to 40 mg, rosuvastatin 10 mg to 40 mg, and simvastatin 20 mg to 80 mg. Doses were titrated to target LDL-C levels specified in the Second Adult Treatment Panel (ATP II) of the National Cholesterol Education Program (NCEP) or to the maximum dosage specified above. Patient-level effectiveness parameters, derived from clinical trial
data and the published literature, included percentage reduction in LDL-C, percentage increase in high-density lipoprotein cholesterol (HDL-C), and achievement of ATP II goal. Resource-use parameters were based on treatment guidelines issued by the Third NCEP Adult Treatment Panel (ATP III) and included statin costs, physician visits, and laboratory tests. Unit costs were based on Medicare reimbursement rates and the wholesale acquisition cost (WAC) of statins. All key parameters were subjected to extensive sensitivity analyses, including Monte Carlo simulation. Details regarding each of these elements of the analysis are given below. The conceptual framework of the model is depicted in Figure 1. The model was constructed using Excel 2002 (Microsoft Corp.) and @Risk 4.5 (Palisade Corp.). The base-case analysis compared the cost-effectiveness of titration to LDL-C goal with each statin in patients with baseline LDL-C ≥ 160 mg/dL and <250 mg/dL. To facilitate comparisons between statins, the incremental cost-effectiveness ratio (ICER) for each drug was the difference in cost between the drug and the next less costly alternative divided by the difference in effectiveness between the drug and the next less costly alternative [19]. Interventions which are more effective and less costly than the alternative (dominant) [19] or which have a lower ICER than the alternative (extended dominance) are preferred [20]. The undominated products comprise the “efficient frontier” [19], which is also referred to in this article as the optimal statin formulary.
Patient Population The analysis employed patient-level data from two multicenter, Phase III, randomized clinical trials. One compared rosuvastatin with atorvastatin [21] and the other compared rosuvastatin with pravastatin and simvastatin [22], both over 52 weeks. These trials had virtually identical inclusion criteria and protocols, and are the only two long-term trials
Figure 1 Model flow—base case. All patients initiated statin therapy on the recommended starting dose and were titrated at their physicians’ discretion until they achieved their ATP II goal or until they reached the highest available dose. Patients who achieved goal continued on maintenance therapy at the current dose. Patients who failed to achieve goal continued on maintenance therapy at the highest available dose.
Benner et al.
620 that compare rosuvastatin to other statins. In the Olsson study [21], 412 patients from 45 centers in northern Europe with LDL-C 160 mg/dL to 249 mg/dL were randomized to rosuvastatin 5 mg, rosuvastatin 10 mg, or atorvastatin 10 mg. After 12 weeks of treatment, dosages could be sequentially doubled up to 80 mg if the patient was not at his/her ATP II LDL-C goal. In the Brown study [22], 477 patients from 43 centers in the United States who met identical inclusion criteria were randomized to rosuvastatin 5 mg, rosuvastatin 10 mg, pravastatin 20 mg, or simvastatin 20 mg. After 12 weeks of fixed-dose therapy, dosages were sequentially doubled to a maximum of 80 mg for rosuvastatin and simvastatin and 40 mg for pravastatin according to investigator discretion if the patient did not meet his/her ATP II LDL-C goal. The 52-week observed data from the Olsson [21] and Brown [22] trials were pooled. Because rosuvastatin was substantially more effective in the European (Olsson [21]) study than in the US (Brown [22]) study, a multivariate linear regression model was fit to the pooled database of rosuvastatin patients from both trials. The main effect for the trial variable (adjusting for clinical and demographic characteristics that were associated with effectiveness) was used to weight the 52-week lipid levels among all Olsson [21] subjects downward to make them similar to those of Brown [22] subjects. The range of effectiveness used in sensitivity analysis was broad enough to cover potential error in the adjustment weights. Subjects who did not complete the 52-week follow-up period were excluded for three reasons. First, the objective of this analysis was to estimate effectiveness and costs in patients treated for at least 1 year. Second, discontinuation rates were low (<20%) and similar across treatment arms, so dropping these patients could be done without biasing the results for any particular drug [21,22]. Third, exclusion of these subjects avoided potentially inaccurate imputation of costs and effectiveness in the base-case analysis. In sensitivity analyses, missing data were imputed for the full intent-to-treat population. Because 10 mg is the recommended starting dose for rosuvastatin [2], and 75% of all rosuvastatin sold is in the form of 10 mg tablets [24], patients randomized to 5 mg were also excluded from this analysis. Based on the characteristics of included trial participants (atorvastatin n = 116; pravastatin n = 95; rosuvastatin n = 202; simvastatin n = 102), the hypothetical cohort in the model was 40% male and had a mean age of 58 years and a mean baseline LDL of 189 mg/dL. Approximately
23% of the patients were considered “high risk” by ATP II criteria; another 35% were at moderately high risk.
Statin Effectiveness Effectiveness measures included the percentage change in LDL-C, the percentage change in HDL-C, and the percentage of patients achieving ATP II LDL-C goal (Table 1). ATP II goals were used for this analysis because comparative 1-year effectiveness data for rosuvastatin is at present only available from trials in which subjects were titrated to ATP II targets. For the base-case analysis, effectiveness of atorvastatin, pravastatin, rosuvastatin and simvastatin was based on measurements at 52 weeks (after titration to goal) in the pooled trial cohort described above. The modest incremental effectiveness in the “base-case” versus “no titration” scenarios may have been because titrations were done after randomization at the physician’s discretion, so patients who were titrated may have been farther from goal or treatment-resistant. Five patients in the trials were titrated to rosuvastatin 80 mg; however, after completion of the clinical program, a decision was made by the manufacturer not to pursue marketing approval of the 80 mg dose. Accordingly, these patients’ 52-week efficacy data were adjusted by carrying forward their last observations at the 40 mg dose. Sensitivity analyses were conducted in which lower and upper 95% confidence limits were used instead of means for effectiveness parameters. These ranges were generally consistent with the results of other long-term clinical trials [2–5,7,25] and product labels. Although fluvastatin and lovastatin were not used in the Olsson [21] and Brown [22] trials, these low-cost statins were included in the CEA to make it more relevant to managed care decision-makers. Effectiveness of these drugs was estimated based on the principle of dose equivalence. Previous research suggests that fluvastatin is half as potent as pravastatin, but the two drugs achieve comparable changes in LDL-C and HDL-C over 52 weeks when given in equipotent doses [26,27]. Thus, because pravastatin was given over the 20 mg to 40 mg dose range in the Brown trial [22], it was assumed that the results would be comparable to fluvastatin over the 40 mg to 80 mg dose range. Similarly, lovastatin 20 mg to 40 mg is approximately equivalent to pravastatin 20 mg to 40 mg [26]. The assumption that fluvastatin 40 mg to 80 mg and lovastatin 20 mg to 40 mg were equivalent to pravastatin 20 mg to 40 mg in terms of LDL-C reduction, HDL-C improvement, and ATP II goal achievement
621
Formulary Decisions in the Statin Class Table 1 Statin effectiveness parameters under base-case and alternative scenarios Statin Atorvastatin 10–80 mg Base-case (95% CI) No titration With ezetimibe Reduced compliance Fluvastatin 40–80 mg Base-case (95% CI) No titration With ezetimibe Reduced compliance Lovastatin 20–40 mg Base-case (95% CI) No titration With ezetimibe Reduced compliance Pravastatin 20–40 mg Base-case (95% CI) No titration With ezetimibe Reduced compliance Rosuvastatin 10–40 mg Base-case (95% CI) No titration With ezetimibe Reduced compliance Simvastatin 20–80 mg Base-case (95% CI) No titration With ezetimibe Reduced compliance
% Reduction in LDL-C
% Increase in HDL-C
% Achieving ATP II goal
38 (36, 40) 36 42 30
0.9 (−1.3, 3.0) 8.5 9.5 —
80 (73, 87) 70 94 52
30 (28, 33) 28 36 24
4.4 (2.1, 6.6) 8.3 9.8 —
60 (50, 70) 56 86 46
30 (28, 33) 28 36 24
4.4 (2.1, 6.6) 8.3 9.8 —
60 (50, 70) 56 86 46
30 (28, 33) 28 36 24
4.4 (2.1, 6.6) 8.3 9.8 —
60 (50, 70) 56 86 46
46 (44, 48) 49 50 36
7.3 (5.4, 9.1) 11.3 11.7 —
87 (83, 92) 89 99 67
37 (34, 39) 37 42 29
6.1 (3.5, 8.7) 9.2 10.1 —
73 (64, 81) 72 91 53
Upper and lower bounds of the 95% CI were used in one-way sensitivity analyses. ATP II, National Cholesterol Education Program, Second Adult Treatment Panel; CI, confidence interval; HDL-C, high-density lipoprotein cholesterol; LDL-C, lowdensity lipoprotein cholesterol; —, not available.
is also consistent with data from long-term trials of those drugs [6,8,25].
Costs Consistent with the managed care perspective, direct medical costs of lipid-lowering therapy were included in this analysis. First-year total cost of treatment was estimated based on statin use, physician office visits, and laboratory monitoring (Table 2). In the base-case analysis, it was assumed that physician visits and lab tests were performed according to ATP III guidelines [28]. The model cohort therefore incurred two physician visits, two lipid panels, one liver function test (LFT), and one creatine kinase (CK) test before initiating therapy, and follow-up visits for titration every 8 weeks until the maximum dose or LDL-C goal was reached. Thereafter, follow-up intervals were reduced to every 6 months. Lipid tests were assumed to occur at each physician visit, and repeat LFT and CK assessments were done at week 8 only. The number of physician visits and lipid tests within the first year of therapy was therefore determined by the number of titrations required to achieve LDL-C goal. Average 2005 Medicare reimbursement rates were used as a measure of the costs of physician vis-
its and laboratory procedures [29,30]. A range of five Common Procedural Terminology, Fourth Edition (CPT-4) codes (99211–99215) for physician office visits were used to estimate this cost. The second highest visit level (second most complicated case) was selected for the base case [31], and the minimum and maximum values in the range were used in sensitivity analyses. CPT-4 codes 80061, 80076, and 82550 were chosen to represent the costs of lipid panels, LFT, and CK, respectively. The minimum and maximum observed values in the range were used in the sensitivity analyses for each parameter. The WAC (80% of average wholesale price [AWP]) for the largest-selling package size as of January 19, 2005 was used to determine the daily cost of each statin. The exception was lovastatin, a multisource generic product, where the average WAC across all manufacturers was used [32]. Because there is great variability in net acquisition cost due to unpublished contractual arrangements and rebates, sensitivity analyses probed a range of WAC less 25% to WAC plus 25% (equivalent to AWP). Threshold analyses identified the percent change in acquisition cost for each product that was required to alter the base-case efficient frontier.
Benner et al.
622 Table 2 Cost parameters (in 2004 US dollars) Parameter Physician visit (CPT 99214) Lipid panel (CPT 80061) Liver function test (CPT 80076) Creatine kinase (CPT 82550) Lipid-lowering drugs (per day) Atorvastatin 10 mg 20/40/80 mg Ezetimibe 10 mg Fluvastatin 40 mg 80 mg Lovastatin 20 mg 40 mg Pravastatin 20 mg 40 mg Rosuvastatin 10/20/40 mg Simvastatin 20/40/80 mg
Base-case
Low value
High value
82.62 18.19 11.02 9.08
21.60 13.69 6.41 8.62
120.14 18.75 11.42 9.10
2.17 3.15
1.63 2.36
2.71 3.94
2.28
1.71
2.85
1.75 2.25
1.31 1.69
2.19 2.81
1.34 2.44
1.01 1.83
1.68 3.05
2.79 4.10
2.09 3.08
3.49 5.13
2.36
1.77
2.95
3.97
2.98
4.96
Compliance with statin therapy was assumed to be 100% in the base-case analysis, because this study pertained only to patients who completed 52 weeks of follow-up, and because differences in adherence between the statins have not been documented. In a separate scenario, the sensitivity of base-case results to reduced compliance was tested.
Adverse Events and Long-Term Outcomes Adverse events were not calculated in the model because available evidence suggests that treatmentlimiting event rates do not differ significantly between the statins [21,22,33]. Moreover, the average cost of adverse events would be low, because most events resolve after discontinuation of the drug [2–8]. Although long-term postmarketing safety data are not yet available for rosuvastatin, cases of rhabdomyolysis are very rare and similar in frequency to the other marketed statins [34]. Also excluded were costs for future clinical outcomes such as myocardial infarction, stroke, coronary artery bypass grafting or percutaneous transluminal coronary angioplasty, even though statin therapy has been shown to reduce the frequency of these procedures [2–8]. Excluding these potential cost offsets is consistent with the short-term time frame of the analysis and gives a more conservative estimate of the cost-effectiveness of statin therapy, particularly among the most effective statins [35].
Sensitivity Analyses and Alternate Scenarios To determine the robustness of base-case findings, alternate assumptions about each variable in the
model were tested. In addition to one-way sensitivity analyses on each variable, a Monte Carlo simulation was used to vary all effectiveness inputs at once [36,37]. The best-fitting continuous probability distribution for each drug’s effectiveness was mathematically identified from the patient-level clinical trial data. Lovastatin and fluvastatin were assigned their own distributions (with the same parameters as pravastatin, as explained earlier), to allow their effectiveness to vary independently from pravastatin in the simulation. Although costs varied with the number of titrations required to reach the goal, statin prices were not varied in the Monte Carlo simulation, as they are constant within a health plan. Values from each probability distribution were randomly selected as the model generated 1000 populations of 10,000 patients each. Costeffectiveness acceptability curves were plotted to display the optimal alternatives as a function of willingness-to-pay for incremental effectiveness [37]. In addition, the impacts of five alternate scenarios that may be reflective of actual practice were also studied. The first scenario assumed that patients were not titrated to goal, but instead completed the year on the initial statin dose. For this analysis, effectiveness at the initial dose was carried forward from the 12-week visit in the clinical trials. In a second scenario, it was assumed that ezetimibe 10 mg was added to statin therapy in lieu of titration for patients who did not achieve LDL-C goal. Patients were assumed to incur an additional visit and lipid panel when initiating ezetimibe, as well as
623
Formulary Decisions in the Statin Class Table 3 1-year and 3-year base-case cost per 1% reduction in LDL-C
Strategy
Average cost ($) 1-year 3-year
Lovastatin Fluvastatin Rosuvastatin Atorvastatin Pravastatin Simvastatin
1280 1242 1326 1401 1778 1955
3120 3177 3476 3676 4846 5280
Incremental cost ($) 1-year 3-year — Dominated 118 Dominated Dominated Dominated
— Dominated 356 Dominated Dominated Dominated
Incremental % ↓ LDL-C 1-year 3-year
Average % ↓ LDL-C 30 30 46 38 30 37
— Dominated 16 Dominated Dominated Dominated
— Dominated 16 Dominated Dominated Dominated
Incremental costeffectiveness ratio ($/1% ↓ LDL-C) 1-year 3-year — Dominated 8 Dominated Dominated Dominated
— Dominated 23 Dominated Dominated Dominated
Cost-effectiveness ratios were calculated before cost and effectiveness estimates were rounded. LDL-C, low-density lipoprotein cholesterol; —, not applicable.
the cost of the drug. The additional effectiveness with ezetimibe add-on therapy was estimated based on efficacy data for patients who have ezetimibe added to an ongoing HMG-CoA therapy regimen which is contained on the product label [38]. The third scenario explored the potential impact of noncompliance on cost-effectiveness. Statin, physician, and laboratory utilization were reduced by 16% based on data from a 12-month study of statin compliance among managed care enrollees [39]. Effectiveness in terms of LDL-C reduction was adjusted downward by 20% in this scenario [40]. Although the Olsson [21] and Brown [22] trials were not powered to detect differences in effectiveness across NCEP ATP II risk subgroups, scenario four examined cost-effectiveness by NCEP ATP II risk group. Finally, the fifth scenario examined costeffectiveness over a 3-year time horizon, consistent with the recommendations of the AMCP Format for Formulary submissions [18].
Results
Base-Case Analysis In the base-case analysis, the mean reductions in LDL-C for rosuvastatin, atorvastatin, and simvastatin were 46%, 38%, and 37%, respectively. Pravastatin, lovastatin, and fluvastatin had mean LDL-C reductions of 30%. Generic lovastatin had
the lowest first-year cost of therapy ($1208), followed by fluvastatin ($1242), rosuvastatin ($1326), atorvastatin ($1401), pravastatin ($1778), and simvastatin ($1955). When the drugs were compared for incremental cost-effectiveness, lovastatin dominated fluvastatin, whereas rosuvastatin dominated atorvastatin, simvastatin, and pravastatin (Tables 3, 4, and 5). The first-year incremental cost of rosuvastatin was thus $118 compared with lovastatin, or $8 per additional 1% reduction in LDLC, $436 per additional patient to ATP II goal, and $41 per additional 1% increase in HDL-C. Atorvastatin 10 mg had a lower acquisition cost than rosuvastatin 10 mg, yet the total first-year cost of rosuvastatin was lower because of greater effectiveness at the starting dose, which prevented titrations in all but 16% of patients. Moreover, the flat price of rosuvastatin minimized the economic impact when patients were titrated. Conversely, 37% of patients treated with atorvastatin required at least one titration. In addition to more frequent physician monitoring and laboratory tests, drug costs increased substantially at higher doses of atorvastatin.
Sensitivity Analyses Base-case results were most sensitive to the acquisition costs of rosuvastatin, atorvastatin, and lovastatin. A decrease in the price of rosuvastatin by 14%
Table 4 1-year and 3-year base-case cost per patient to ATP II goal*
Strategy Lovastatin Fluvastatin Rosuvastatin Atorvastatin Pravastatin Simvastatin
Average cost ($) 1-year 3-year 12,080,450 12,424,554 13,264,305 14,005,258 17,780,272 19,547,992
31,197,003 31,767,946 34,756,905 36,763,638 48,457,949 52,801,642
Incremental cost ($) 1-year 3-year — Dominated 1,183,855 Dominated Dominated Dominated
— Dominated 3,559,902 Dominated Dominated Dominated
Average patients to ATP II goal 6,000 6,000 8,713 8,017 6,000 7,255
*Assuming 10,000 patients treated with each statin. Cost-effectiveness ratios were calculated before cost and effectiveness estimates were rounded. ATP II, National Cholesterol Education Program, Second Adult Treatment Panel; —, not applicable.
Incremental patients to ATP II goal 1-year 3-year — Dominated 2,713 Dominated Dominated Dominated
— Dominated 2,713 Dominated Dominated Dominated
Incremental costeffectiveness ratio ($/patient to ATP II goal) 1-year 3-year — Dominated 436 Dominated Dominated Dominated
— Dominated 1,312 Dominated Dominated Dominated
Benner et al.
624 Table 5 1-year and 3-year base-case cost per 1% increase in HDL-C
Strategy Lovastatin Fluvastatin Rosuvastatin Atorvastatin Pravastatin Simvastatin
Average cost ($) 1-year 3-year 1208 1242 1326 1401 1778 1955
3120 3177 3476 3676 4846 5280
Incremental cost ($) 1-year 3-year — Dominated 118 Dominated Dominated Dominated
— Dominated 356 Dominated Dominated Dominated
Incremental % ↑ HDL-C 1-year 3-year
Average % ↑ HDL-C 4.4 4.4 7.3 0.9 4.4 6.1
— Dominated 2.9 Dominated Dominated Dominated
— Dominated 2.9 Dominated Dominated Dominated
Incremental costeffectiveness ratio ($/1% ↑ HDL-C) 1-year 3-year — Dominated 41 Dominated Dominated Dominated
— Dominated 122 Dominated Dominated Dominated
Cost-effectiveness ratios were calculated before cost and effectiveness estimates were rounded. HDL-C, high-density lipoprotein cholesterol; —, not applicable.
or more made it the least costly and most effective alternative, dominating all other statins. Conversely, for prices up to 30% greater than its WAC, rosuvastatin remained the only branded statin on the efficient frontier. Nevertheless, the ICER at WAC plus 30% was $24 per additional 1% decrease in LDL (three times the base case ICER), compared with lovastatin. When the price of atorvastatin was discounted 15% below WAC, it joined the efficient frontier with an ICER of $7 per additional 1% decrease in LDL compared with lovastatin. The ICER for rosuvastatin in that scenario was $8 compared with atorvastatin. When atorvastatin was discounted 24% or more, it became the least costly statin, and the ICER for rosuvastatin was $18 per additional 1% decrease in LDL. When the average price of lovastatin was replaced with the lowest available price (36% less than the base-case price), its total cost in year 1 was $956, and the ICER for rosuvastatin was $24 per additional 1% decrease in LDL compared with lovastatin. Pravastatin and simvastatin remained dominated until their base-case prices were discounted 45% and 49%, respectively. Results were moderately sensitive to assumptions about HDL-C improvement. Moderate increases in the base-case HDL-C benefits of simvastatin and fluvastatin were sufficient for these products to avoid being dominated. For example, if fluvastatin improved HDL-C by 6.6%, it joined the efficient frontier, and the ICER for rosuvastatin increased threefold ($123 per additional 1% increase in HDLC compared with fluvastatin). When simvastatin was assumed to increase HDL-C by 8.7%, it joined the efficient frontier, but its relatively high price yielded an ICER of $443 per additional 1% improvement in HDL-C, compared with rosuvastatin. Results were insensitive to the cost of physician visits and lab procedures, as well as to variations in effectiveness over the 95% confidence limits for LDL-C reduction. At its upper confidence limit for ATP II goal achievement, atorvastatin was the most
effective statin. Nevertheless, its higher price and small incremental benefit relative to rosuvastatin yielded a high ICER ($27,313 per additional patient to goal versus rosuvastatin). Results did not differ substantially from the base case when costs and benefits were imputed for patients who discontinued therapy during the study year. Figure 2 provides the cost-effectiveness acceptability frontier for each outcome of interest, based on 1000 simulated populations of 10,000 patients each. In all cases, the frontier included only lovastatin and rosuvastatin. The intersection of the lines illustrates the willingness-to-pay threshold where the optimal drug of choice changes from lovastatin to rosuvastatin. These willingness-to-pay thresholds are approximately $8, $434, and $40 for a 1% reduction in LDL-C, one additional patient to ATP II goal, and a 1% increase in HDL-C, respectively.
Alternative Scenarios When patients were assumed to remain at their respective starting doses and 12-week effectiveness persisted for the full year, lovastatin remained the least costly alternative ($933), followed by fluvastatin ($1083), atorvastatin ($1236), rosuvastatin ($1305), pravastatin ($1463), and simvastatin ($1894). Although atorvastatin was less costly than rosuvastatin in this scenario, atorvastatin was still dominated in the extended sense because rosuvastatin had a lower incremental cost per additional unit of benefit compared with lovastatin. When patients received add-on therapy with ezetimibe instead of statin titration, lovastatin remained the least costly strategy and rosuvastatin was the most effective, with an ICER of $10 per additional 1% reduction in LDL-C. Rosuvastatin dominated fluvastatin, atorvastatin, pravastatin, and simvastatin. Similarly, when resource use and effectiveness were adjusted for compliance rates representative of actual practice, the results were virtually identical to the base-case.
625
Formulary Decisions in the Statin Class A. Cost-effectiveness acceptability curves for LDL-C reduction 100.00 Rosuvastatin
% of Time Optimal
80.00 60.00 Threshold = $7.60 40.00 20.00 Lovastatin 0.00 $0.00
$5.00 $10.00 Willingness To Pay
$15.00
B. Cost-effectiveness acceptability curves for HDL-C increase 100.00
% of Time Optimal
Rosuvastatin 80.00 60.00 Threshold = $40.32 40.00 20.00
Discussion
Lovastatin 0.00 $0.00
$20.00
$40.00 $60.00 Willingness To Pay
$80.00
C. Cost-effectiveness acceptability curves for ATP II goal achievement 100.00 Rosuvastatin
% of Time Optimal
80.00
60.00 Threshold = $434.06 40.00
20.00 Lovastatin 0.00 $0.00
$200.00
$400.00 $600.00 Willingness To Pay
ditional 1% reduction in LDL for rosuvastatin versus lovastatin was $6, $10, and $7 in the high-risk, moderately high risk, and low-risk groups, respectively. These ratios reflect the fact that LDL reductions were greatest among the high-risk patients (51% for rosuvastatin and 32% for lovastatin), followed by low-risk patients (46% and 29%), and moderately high risk patients (43% and 31%). Finally, when the base-case scenario was evaluated using a 3-year time horizon, total costs increased to reflect longer-term statin use (Tables 3, 4, and 5). Nevertheless, effectiveness was the same as in the 1-year analysis because, under recommended monitoring and titration intervals, all titrations occur within the first year of treatment. Thus, the ICERs in the 3-year analysis may be interpreted as the cost to maintain a given level of effectiveness for 3 years.
$800.00
Figure 2 Results of Monte Carlo simulation. (A) Cost-effectiveness acceptability curves for LDL-C reduction. The threshold is interpreted as the point at which each therapy is optimal 50% of the time based on an expressed willingness to pay for each incremental 1% reduction in LDL-C. (B) Cost-effectiveness acceptability curves for HDL-C increase. The threshold is interpreted as the point at which each therapy is optimal 50% of the time based on an expressed willingness to pay for each incremental 1% increase in HDL-C. (C) Costeffectiveness acceptability curves for ATP II goal achievement. The threshold is interpreted as the point at which each therapy is optimal 50% of the time based on an expressed willingness to pay for each incremental patient to ATP II goal.
In the subanalysis of cost-effectiveness by NCEP ATP II risk group, the base-case results remained relatively stable, but the cost-effectiveness ratio for rosuvastatin was most favorable for patients at high risk of coronary events. The incremental cost per ad-
This CEA compared currently available statins in patients with dyslipidemia from the perspective of a managed care payer. When patients were titrated to ATP II goals, the most effective strategy was rosuvastatin over the 10 mg to 40 mg dose range. Atorvastatin, simvastatin, and pravastatin were less effective and more costly than rosuvastatin. Lovastatin over the 20 mg to 40 mg range was the least costly strategy, followed by fluvastatin over the 40 mg to 80 mg range. Compared with lovastatin, the incremental cost of rosuvastatin was $118 per patient in the initial year of therapy, or $8 per additional 1% reduction in LDL-C, $41 per additional 1% increase in HDL-C, and $436 per additional patient to ATP II goal. These ratios increased by as much as three times as the price of generic lovastatin was reduced from the average of all available manufacturers to the lowest published price. These findings have potentially important implications for managed care decision-makers. Under the base-case and each alternative scenario, atorvastatin, pravastatin, and simvastatin were always dominated. Thus, depending on actual acquisition prices, payers may achieve substantial cost savings and greater effectiveness by using rosuvastatin instead of these agents. In the Monte Carlo simulation of 1000 populations, the cost-effectiveness acceptability curves favored generic lovastatin and rosuvastatin as the optimal formulary. This conclusion was consistent with that of Morrison and Glassberg, who asserted that a cost-effective approach to lipid-lowering therapy would be to treat low-risk patients and those with lower base-
626 line LDL-C levels with the least costly statin, while using the most effective drug in high-risk patients and those with high baseline LDL-C levels [41]. The present study extends those theoretical findings based on currently available pricing and clinical data. Moreover, the results of this analysis can be used to calculate the trade-offs inherent in selecting a different formulary. For example, a payer might consider placing generic lovastatin and simvastatin on formulary until long-term outcomes and safety data become available for rosuvastatin. Assuming 10,000 patients treated with each product, in year 1 the choice of simvastatin would cost about $6.3 million more than rosuvastatin, and 1458 fewer patients would achieve ATP II LDL-C goals. There remains an important question of whether the additional cost of rosuvastatin compared with lovastatin ($436 per additional patient to LDL-C goal) constitutes good value for money [42]. To put this in a decision-making context, the analysis may be recalculated in the absence of rosuvastatin. The next most effective alternative is atorvastatin, which is currently on many formularies. Under base-case assumptions, atorvastatin, in the absence of rosuvastatin, has an incremental cost per patient to goal of $954, compared with lovastatin. Thus, rosuvastatin is able to get more patients to goal than atorvastatin, and at a 54% lower incremental cost per patient to goal. This analysis was unique in its analytic approach, because it was neither purely trial-based nor purely model-based. Typical trial-based analyses are subject to the limitations of protocol-driven resource use, which does not apply to real-world populations. Conversely, purely model-based analyses require that multiple sources of data be synthesized and extrapolated to common endpoints in ways that enhance external validity at the expense of internal validity. Given that no long-term outcomes or safety studies are yet available for rosuvastatin or ezetimibe, we developed a simple model based on clinical guidelines for the management of dyslipidemia patients, but limited the base-case time horizon and effectiveness endpoints to those used in the available clinical trials. This approach thus placed the available clinical data in the context of a real-world formulary decision, thereby balancing the tradeoffs between internal and external validity. It may well be a useful method for evaluating new products in other therapeutic areas. These findings should be interpreted in light of some limitations. Because the objective was to identify the optimal combination of statins for a managed care formulary, several nonstatin medications
Benner et al. for dyslipidemia (e.g., niacin, fibrates, and bile sequestrants) were excluded from analysis. These products generally have a secondary role in therapy, are relatively inexpensive, and were not expected to influence the relative cost-effectiveness of products in the statin class. Their exclusion from this analysis should not be construed to mean that they are not cost-effective agents for modifying serum lipids. The effectiveness of atorvastatin, pravastatin, rosuvastatin, and simvastatin were estimated based on data from two Phase III trials. Fluvastatin and lovastatin were assumed to be equivalent to pravastatin over the dose ranges specified, because long-term trials have not compared fluvastatin and lovastatin with rosuvastatin. Although the resulting estimates for each drug are consistent with numerous other trials and product labels, and although these estimates were varied in sensitivity analyses, the results may not apply to patient populations with different characteristics than those in the Olsson [21] and Brown [22] trials. The subgroup analysis by NCEP ATP II risk group should be interpreted with caution, as the trials were not powered to detect differences in effectiveness across risk strata. Costs were estimated using published WAC for drugs and Medicare reimbursement rates for physician visits and laboratory procedures. These costs may not reflect the actual costs negotiated by managed care organizations, but wide variation in these parameters were probed to find relevant thresholds that changed the optimal formulary. Costs associated with adverse drug events and nonstudy medications were excluded, but these were not expected to differ across treatment groups, and therefore would not have affected the incremental analysis. This analysis had a 1- to 3-year time horizon, and therefore examined surrogate endpoints instead of long-term outcomes such as cost per life-year gained. Although long-term head-to-head outcomes studies comparing the agents in this analysis would be a preferable data source [43], decisions must be made based on currently available data. As such, this model informs payers of the trade-offs inherent in covering one statin (or combination of statins) over another. Moreover, studying surrogate endpoints in this analysis yielded conservative estimates of the incremental effectiveness of rosuvastatin. Estimating long-term events prevented and associated cost offsets based on recent clinical trial data [35] would increase the incremental effectiveness and decrease the incremental cost of rosuvastatin. Finally, achievement of ATP III LDL-C goals could not be measured because patients were titrated to ATP II goals in the Olsson [23] and Brown [24] tri-
627
Formulary Decisions in the Statin Class als. Nevertheless, a recent head-to-head trial of ATP III goal achievement after 6 weeks of treatment with rosuvastatin, atorvastatin, simvastatin, or pravastatin resulted in incrementally higher goal achievement with rosuvastatin [44]. The magnitude of the incremental effectiveness was similar to the inputs employed in the present analysis. In conclusion, the findings of this analysis indicate that rosuvastatin is less costly and more effective than atorvastatin, pravastatin, and simvastatin over the dose ranges for which comparative data are available. Compared with fluvastatin and pravastatin, generic lovastatin is cost-saving and about as effective. Therefore, generic lovastatin and rosuvastatin comprise the optimal two-statin formulary. Compared with generic lovastatin, rosuvastatin offers substantial additional LDL-C reduction, HDL-C improvement, and NCEP goal achievement at a reasonable additional cost, depending on the payer’s willingness to pay for these benefits. Formulary decisions based on these results should be revisited periodically, as new pricing, outcomes and safety data become available.
7
8
9
10
11
12 Source of financial support: This research was sponsored by AstraZeneca.
13
References 1 American Heart Association. 2004 Heart and Stroke Statistical Update. Dallas, TX: American Heart Association, 2003. 2 The Scandinavian Simvastatin Survival Study Group. Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S). Lancet 1994;344:1383–9. 3 Shepherd J, Cobbe SM, Ford I, et al. for the West of Scotland Coronary Prevention Study Group. Prevention of coronary heart disease in men with hypercholesterolemia. N Engl J Med 1995;333: 1301–7. 4 Sacks FM, Pfeffer MA, Moye LA, et al. for the Cholesterol and Recurrent Events Trial Investigators. The effect of pravastatin on coronary events after myocardial infarction in patients with average cholesterol levels. N Engl J Med 1996;335: 1001–9. 5 The Long-Term Intervention with Pravastatin in Ischaemic Disease (LIPID) Study Group. Prevention of cardiovascular events and death with pravastatin in patients with coronary heart disease and a broad range of initial cholesterol levels. N Engl J Med 1998;339:1349–57. 6 Downs JR, Clearfield M, Weis S, et al. for the AFCAPS/TexCAPS Research Group. Primary pre-
14
15
16
17
18
vention of acute coronary events with lovastatin in men and women with average cholesterol levels: results of AFCAPS/TexCAPS. JAMA 1998;279: 1615–22. Heart Protection Study Collaborative Group. MRC/BHF heart protection study of cholesterol lowering with simvistatin in 20,536 high-risk individuals: a randomised placebo-controlled trial. Lancet 2002;360:7–22. Serruys PWJC, de Feyter P, Macaya C, et al. Fluvastatin for prevention of cardiac events following successful first percutaneous coronary intervention. JAMA 2002;287:3215–22. Jackevicius CA, Anderson GM, Leiter L, Tu JV. Use of the statins in patients after acute myocardial infarction: does evidence change practice? Arch Intern Med 2001;161:183–8. Lemaitre RN, Furberg CD, Newman AB, et al. Time trends in the use of cholesterol-lowering agents in older adults. Arch Intern Med 1998;158: 1761–8. Hoerger TJ, Bala MV, Bray JW, et al. Treatment patterns and distribution of low-density lipoprotein cholesterol levels in treatment-eligible United States adults. Am J Cardiol 1998;82:61–5. Stafford RS, Blumenthal D, Pasternak RC. Variations in cholesterol management practices of U.S. physicians. J Am Coll Cardiol 1997;29:139–46. Fonarow GC, French WJ, Parsons LS, et al. Use of lipid-lowering medications at discharge in patients with acute myocardial infarction: data from the National Registry of Myocardial Infarction 3. Circulation 2001;103:38–44. Pearson TA, Laurora I, Chu H, Kafonek S. The lipid treatment assessment project (L-TAP): a multicenter survey to evaluate the percentages of dyslipidemic patients receiving lipid-lowering therapy and achieving low-density lipoprotein cholesterol goals. Arch Intern Med 2000;160:459–67. Foley KA, Simpson RJ, Crouse JR, et al. Effectiveness of statin titration on low-density lipoprotein cholesterol goal attainment in patients at high risk of atherogenic events. Am J Cardiol 2003;92:79– 81. Ford ES, Mokdad AH, Giles WH, Mensah GA. Serum total cholesterol concentrations and awareness, treatment, and control of hypercholesterolemia among US adults: findings from the National Health and Nutrition Examination Survey, 1999– 2000. Circulation 2003;107:2185–9. McKenney JM, Jones PH, Adamczyk MA, et al. Comparison of the efficacy of rosuvastatin vs. atorvastatin, simvastatin, and pravastatin in achieving lipid goals: results from the STELLAR trial. Curr Med Res Opin 2003;19:689–98. Academy of Managed Care Pharmacy. Format for Formulary Submissions, Version 2.0. Alexandria, VA: Academy of Managed Care Pharmacy, 2002.
Benner et al.
628 19 Garber AM, Weinstein MC, Torrance GW, Kamlet MS. Theoretical foundations of cost-effectiveness analysis. In: Gold MR, Siegel JE, Russel LB, Weinstein MC, eds., Cost-Effectiveness in Health and Medicine. New York: Oxford University Press, 1996. 20 Siegel JE, Weinstein MC, Torrance GW. Reporting cost-effectiveness studies and results. In: Gold MR, Siegel JE, Russel LB, Weinstein MC, eds., CostEffectiveness in Health and Medicine. New York: Oxford University Press, 1996. 21 Olsson AG, Istad H, Luurila O, et al. Effects of rosuvastatin and atorvastatin compared over 52 weeks of treatment in patients with hypercholesterolemia. Am Heart J 2002;144:1044–51. 22 Brown WV, Bays HE, Hassman DR, et al. Efficacy and safety of rosuvastatin compared with pravastatin and simvastatin in patients with hypercholesterolemia: a randomized, double-blind, 52-week trial. Am Heart J 2002;144:1036–43. 23 AstraZeneca LP. Crestor® (rosuvastatin calcium) prescribing information. August 2003. 24 First Data Bank. Data on file, AstraZeneca. 25 Davidson M, McKenney J, Stein E, et al. Comparison of one-year efficacy and safety of atorvastatin vs. lovastatin in primary hypercholesterolemia. Am J Cardiol 1997;79:1475–81. 26 Roberts WC. The rule of 5 and the rule of 7 in lipid-lowering by statin drugs [Editorial]. Am J Cardiol 1997;80:106–7. 27 Andrews TC, Ballantyne CM, Hsia JA, Kramer JH. Achieving and maintaining National Cholesterol Education Program low-density lipoprotein cholesterol goals with five statins. Am J Med 2001;111: 185–91. 28 Third Report of the Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III). National Cholesterol Education Program, National Heart Lung and Blood Institute November 19, 2002. WWW document: Available from: http://www.nhlbi.nih.gov/guidelines/cholesterol/ atp3_rpt.htm [Accessed June 4, 2004]. 29 2004 National Physician Fee Schedule Relative Value File. WWW document: Available from: http://www.cms.hhs.gov/providers/pufdownload/ rvudown.asp [Accessed June 4, 2004]. 30 2004 Clinical Diagnostic Laboratory Fee Schedule. WWW document: Available from: http://www.
31
32 33 34 35
36
37
38 39 40
41 42 43 44
cms.hhs.gov/providers/pufdownload/clfdown.asp [Accessed June 4, 2004]. Ganz DA, Kuntz KA, Jacobson GA, Avorn J. Costeffectiveness of 3-hydroxy-3-methylglutaryl Coenzyme A reductase inhibitor therapy in older patients with myocardial infarction. Ann Intern Med 2000;132:780–7. First DataBank National Drug Data File. Data on file, Astra Zeneca. Thompson PD, Clarkson P, Karas RH. Statinassociated myopathy. JAMA 2003;289:1681–90. Olsson GO. Safety and efficacy of rosuvastatin [Letter]. Lancet 2004;364:135. Cannon CP, Braunwald E, McCabe CH, et al. Intensive versus moderate lipid lowering with statins after acute coronary syndromes. N Engl J Med 2004;350:1495–504. Doubilet P, Begg CB, Weinstein MC, et al. Probabilistic sensitivity analysis using Monte Carlo simulation: a practical approach. Med Decis Making 1985;5:157–77. Briggs AH, O’Brien BJ, Blackhouse G. Thinking outside the box: recent advances in the analysis and presentation of uncertainty in costeffectiveness studies. Annu Rev Public Health 2002;23:377–401. Zetia® (ezetimibe) Prescribing Information. March 2003, Merck/Schering-Plough Pharmaceuticals, Inc., Table 2. Fitterman LK, Williams SA. A descriptive review of statin utilization patterns. Am Heart J 2002;144: 905–6. Frolkis JP, Pearce GL, Nambi V, et al. Statins do not meet expectation for lowering low-density lipoprotein cholesterol levels when used in clinical practice. Am J Med 2002;113:625–9. Morrison A, Glassberg H. Determinants of the cost-effectiveness of statins. J Manag Care Pharm 2003;9:544–51. Ohsfeldt RL. Challenges in evaluating the costeffectiveness of statins [Letter]. J Manag Care Pharm 2004;10:81–2. Hay JW. Evaluating the cost-effectiveness of statins [Letter]. J Manag Care Pharm 2004;10:79–81. Jones PH, Davidson MH, Stein EA, et al. Comparison of the efficacy and safety of rosuvastatin versus atorvastatin, simvastatin, and pravastatin across doses (STELLAR Trial). Am J Cardiol 2003;93:152–60.